https://github.com/sayansomya/defect-detection-in-3d-printing
Bachelors Thesis Project for Defect Detection
https://github.com/sayansomya/defect-detection-in-3d-printing
Last synced: 11 months ago
JSON representation
Bachelors Thesis Project for Defect Detection
- Host: GitHub
- URL: https://github.com/sayansomya/defect-detection-in-3d-printing
- Owner: SayanSomya
- Created: 2024-11-18T20:28:17.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-11-18T21:32:02.000Z (over 1 year ago)
- Last Synced: 2025-03-29T06:47:20.207Z (over 1 year ago)
- Language: Jupyter Notebook
- Size: 36.1 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Defect Detection in 3D Printing
Bachelor's Thesis Project
Welcome to the official repository for Defect Detection in 3D Printing, a Bachelor's Thesis project focused on leveraging advanced object detection models (YOLOv5 and YOLOv11) to identify defects in 3D-printed objects. This project combines computer vision, machine learning, and domain-specific problem-solving to enhance the quality control process in additive manufacturing.
## Project Overview
Additive manufacturing, or 3D printing, has revolutionized how we create objects by enabling complex, highly customizable designs. However, the quality assurance of 3D-printed components remains a significant challenge due to defects that may occur during printing. This project addresses this challenge by implementing state-of-the-art object detection models to automate defect detection, thereby reducing the reliance on manual inspection.
Key Features:
1. Automated Defect Detection: Identify common defects such as layer shifts, stringing, and under-extrusion.
2. YOLO-based Object Detection Models: Evaluate and compare YOLOv5 and YOLOv11 for detecting anomalies in 3D prints.
3. Custom Dataset: Built a comprehensive dataset containing annotated defect images for model training and testing.
4. Performance Optimization: Focused on precision, recall, and inference time to ensure practical usability in real-time applications.
## Methodology
1. Dataset Preparation: Labeled the dataset with bounding boxes for precise defect localization & performed image augmentation to improve model generalization.
2. Model Training: Implemented YOLOv5 and YOLOv11 in Google Collab and optimized hyperparameters for better accuracy and inference speed.
3. Evaluation: Compared YOLOv5 and YOLOv11 based on precision, recall, F1-score, and mAP (Mean Average Precision).
## Future Work
1. Extend the dataset with more defect types and variations.
2. Explore other state-of-the-art object detection architectures like YOLOv11 and DETR.
3. Integrate the detection system into a 3D printing pipeline for real-time defect monitoring.